Complements, relatives, and nominal properties Anna Roussou ([email protected]) University of Patras GIST5:“Generalizing relative strategies” University of Ghent, 22-23.03.2012

Aim: to discuss the similarities between relative and complement clauses, focusing on the Greek complementizer pu.

Goal: to show that a uniform analysis of relatives and complements with pu is possible, drawing on the nominal properties of complementizers.

1. Introduction The element pu introduces relative clauses, and a subset of complement clauses:

(1) Sinandisa ton fititi pu/o opios pire to vravio. Relative clause (RC) met-1s the student that/the which got-3s the prize “I met the student that got the prize.” (2) Lipame pu efijes. Complement clause (CC) be.sorry-1s the left-2s “I’m sorry that you left.” (3) (ton) idha (ton fititi) pu efevje. Pseudorelative (PRC) him saw-1s the student that was.leaving-3s “I saw the student leaving”

• Pu developed from the relative locative adverb opou > pu; a generalized relativizer. (4) to meros pu pigame (the place where/that we went), o logos pu irthame (the reason why/that we came), etc.

• Complement clauses: with emotive predicates, a sub-set of factives, as in (2). Factive complements with pu are strong islands (Roussou 1994, Varlokosta 1994):

(5) a. *?Pjoni metanioses pu sinandises ti? Argument extraction who regretted-2s that met-2s “?Who did you regret that you met?” b. *Posi metanioses pu taksidepses stin Athina ti? Adjunct extraction how regretted-2s that travelled-2s to-the Athens “*How did you regretted that you travelled to Athens?” (e.g. by train)

• Relative clauses are also strong islands (CNPC, Ross 1967)

(6) a. *Tii sinandises ton fititi pu pire ti? Argument extraction what met-2s the student that got-3s “*What did you meet the student that got?” b. *Pote sinandises ton fititi pu pire to vravio ti? Argument extraction when met-2s the student that got-3s the prize “*When did you meet the student that got the prize?” (*= embedded)

[1]

• Questions: a) Is pu in relative and complement clauses the same element? b) Is it the same as interrogative pu (‘where’)? c) If so, what is the inherent property of pu that allows for this distribution and how are the various readings (syntactically) distinguished? d) Are complement clauses hidden relatives? Is this the right question to address and if so, how?

2. One, two, how many pu? 2.1 Background • The dual status of pu: a) a relativizer, b) a complementizer. • Pu as a relativizer: in both restrictive and non-restrictive relative clauses. In restrictive relatives, resumption inside the pu-clause is required with indirect arguments (7b); in other cases pu may stand for a PP (7c) (see Theophanopoulou 1985, Alexopoulou 2006):

(7) a. Sinandisa ton fititi ston opio edhoses to vravio. met-1s the student to-the whom gave-2s the prize b. Sinandisa ton fititi pu *(tu) edhoses to vravio. met-1s the student that he.gen gave-2s the prize “I met the student to whom/*that you gave the prize” c. Ine i triti dulia pu/ ja tin opia kano etisi. (Alexopoulou 2006:99) is the third job that/for the which make-1s application “It’s the third job that I apply for/ for which I apply.”

• Pu in complement clauses: pu as the equivalent of the definite determiner at the clausal level (Christidis 1986, adapted by Roussou 1994, 2010, Varlokosta 1994 a.o.). a) pu is [+def] or [discourse familiar] and resides in C (Roussou 1994); b) a complex structure with movement of pu from C to D[+def] (Roussou & Roberts 2001); c) pu-clauses stand in a paratactic relation with an empty pronominal in the main clause (apposition) (Varlokosta 1994).

(8a) CP (8b) DP 3 3 pu IP pui CP 3 ti IP

(8c) [….[VP V proi] … [CP pu-clause]i ]

(i) Arguments in favour of (8a-b): pu-clauses cannot be nominalized. (9) [To oti/*pu efije] me ekseplikse. the that left-3s me surprised-3s “It surprised me that he left/The fact that he left surprised me”.

(ii) The structure in (c) renders pu-complements more like relatives. Arguably that can hold for (b), under the raising analysis of relative clauses (Kayne 1994).

• A third (?) pu: interrogative (stressed): (10) a. Pú pas? where go-2s “Where are you going?”

[2]

Wh-words are standardly analyzed as phrasal, occupying Spec,CP:

b. CP 3 pu CP 3 C IP

2.2 Pu: a locative pronoun • Roussou 2010: complementizers oti (that), pu (that/‘where’), an (if/whether) are nominal elements (see also Manzini & Savoia 2003) • Relative/interrogative pronouns as complementizers: que, che in Romance (Manzini 2010). • Que/che/that as complementizers: they bind a propositional variable; que/che/that as pronouns: they bind an individual variable (see above references; on English that, see Roberts & Roussou 2003). • A wh-word as a complementizer in English as well: how (Legate 2010)

(10) They told me how/that the tooth fairy doesn’t really exist.

• Pos ‘how’ in Greek may also be used instead of oti (there is no systematic study of their potential distributional differences; oti is more formal).

(11) Mu ipan pos/oti tha fighume. me told how/that fut. leave-1p “They told me that we would leave.”

• Going back to pu: if pu is definite, how can it appear in relative clauses headed by an indefinite head, or embed what looks like a subjunctive (na-clause)?

(12) a. Sinandisa enan/kapjon fititi pu ixe pari to vravio to 2010. met-1s a/ some student that had-3s gotten the prize the 2010 “I met a/some student that had gotten the prize in 2010.” b. Dhen sinandisa kanenan pu na milai kala Italika. not met-1s anyone that prt speak-3s good Italian “I didn’t meet anyone who could speak Italian well.”

If not definite, what? • Pu is a relative/interrogative pronoun (the interrogative reading is marked by stress), which is characterized as [locative]. • As a pronoun (relative/interrogative) it binds an individual variable corresponding to a locative associated with the predicate (perhaps as part of its event structure). • Does pu bind a propositional variable in complement clauses, as argued for oti and an (Roussou 2010). In order to provide an answer to this question, we need to re-consider the predicates that select for pu.

Before that: Bulgarian deto (Krapova 2010, hers (1)): (13) a. Tova e čovekăt, deto (go) snimax včera. this is man-the that him.ClAcc photographed-1s yesterday “This is the man that I photographed yesterday.”

[3]

b. Săžaljavam, deto ne možax da dojda. regret-1s that not could-1s mod come-1s “I regret that I couldn’t come.”

Also used with an indefinite head (fn. 49): (14) Njama čovek, deto ne revnuna. has-not person that not is-jealous “There is nobody who is not jealous”

The origin of deto: the locative kădeto ‘where’> gdeto/deto. (The other complementizer, similar to Greek oti, is če).

In short: • If definiteness is not the relevant property underlying pu (and deto), then its presence in relatives headed by an indefinite is not a problem. • We now need to address the role of pu in complement clauses, and (re-)examine the claim that pu is a factive C. If yes, then it somehow differs from the relative one, and if no, we need to account for the factive reading associated with these complements.

3. Factives, relatives and the role of pu 3.1 Factives and referentiality • Complementizer pu as a marker of factivity (Christidis 1981, 1986). However, only a sub- class of factive predicates, that of emotives, select pu. Factives like paradhexome (admit) or antilamvanome (realize) take an oti-complement:

(14) a. O Janis (dhen) paradextike oti eklepse ta lefta. the John not admitted-3s that stole-3s the money “John admitted (/didn’t admit) that he had stolen the money.” b. O Janis (dhen) antilifthike oti tu eklepsan ta lefta. the John not realized-3s that his stole-3s the money “John realized (/didn’t realize) that someone had stolen his money.”

Some emotives may alternate between oti and pu, with different readings:

(15) a. O Janis xerete pu/*oti efijes. the John be.glad-3s that left-2s “John is glad that you came.” b. O Janis anisixi pu/oti efijes. the John worry-3s that left-2s “John worries about the fact/that you left.”

(15b) with pu: the propositional content of the embedded clause is true (presupposition); with oti it is not (it may or not be the case that you left).

In some cases, pu may be triggered when the selecting predicate is focused, as in (16):

(16) a. Su IPA pu efije o Janis (!) you tell-1s that left the John “I did tell you that John left!” b. EMATHES pu efije o Janis?

[4]

learn-2s that left-2s the John “Did you learn (hear) that John left?”

A ‘factive’ reading under focus (see also Kalluli 2008).

• Factivity with emotives: the lexical properties of the predicates along with pu. There is no one-to-one correspondence between factivity and pu: the former can appear with oti as well, while pu also appears in non-factive clauses, i.e. relatives.

• Another route: Clarify the notions: ‘factivity’, ‘presupposition’, ‘background/familiar information’ (i) de Cuba and Ürögdi (2010): factivity is associated with the lexico-semantic properties of the selecting predicates (emotives are a sub-class of factives). (ii) Factive complements are referential (Haegeman and Ürögdi 2010)

• Advantages: (a) maintain a single pu (relatives, interrogatives, complements of emotives), (b) derive the associated ‘definiteness’ property through ‘referentiality’ (definite descriptions are referential), which is, however, not inherent to pu by arises as a property of the selecting predicate.

• So what is the role of pu in complement clauses (as opposed to oti)? It assigns a reference to the clause it embeds, i.e. it does not bind a propositional variable, since the embedded clause is interpreted as true (no true/false disjunction).

(17) The three complementizers: (a) oti (that): binds a propositional variable (set of propositions) (b) an (if): binds a propositional variable – also polarity sensitive (see (18)) (c) pu (that): assigns reference (single proposition)

(18) a. Rotisa an eftase o Petros. asked-1s if arrived-3s the Peter “I asked if Peter arrived.” b. Dhen proseksa an eftase o Petros. not noticed-1s if arrived-3s the Peter “I didn’t notice whether Peter arrived.”

Notice that proseksa can also carry a factive interpretation, when the oti-clause is associated with an accusative clitic, i.e. dhen to proseksa oti eftase o Petros (he did arrive, but I haven’t noticed it).

• Syntax plays an important role in affecting the selecting properties and interpretation of the predicate.

3.2 The structure of pu-complements and implications for their islandhood • Locality attributed to a +definiteness feature on C, or on a D head to which pu moves, or to a paratactic construction (basically adjunction). • However, ungrammaticality is not as strong as the one we get out of adjunct clauses, or out of relative clauses. In the latter case, the presence of another gap inside the relative clause, obviously deteriorates grammaticality.

[5]

(19) a. *?Pjoni stenoxorjese pu apelisan ti? who be.upset-2s that fired-3p “?Who are you upset that they fired?” b. *Pjoni stenoxorjese epidhi apelisan ti? who be.upset-2s because fired-3p “*Who are you upset because they fired?”

• The syntax of emotives is a bit more complex and not uniform: verbs like stenoxorjeme, lipame, xerome, anisixo are subject experiencer verbs (anisixo can also be an object experience) that take a a ‘Subject Matter’ or ‘Target of Emotion’ object (as argued for by Pesetsky 1995).

(20) a. Stemoxorjeme/anisixo me ton Petro (Subject Matter of Emotion) be.upset-1s/worry-1s with the Peter “I’m upset/worried about Peter” b. Stemoxorjeme/anisixo jia ton Petro (Target of Emotion) be.upset-1s/worry-1s for the Peter “I’m upset/worried for Peter”

In both cases the object is realized as a PP, i.e. ‘oblique’.

• The pu-complement also expresses the ‘Target of Emotion’, and can be considered as corresponding to an ‘oblique’ argument.

• Similar patterns in Bulgarian (Krapova 2010, her (59)): (21) Săžaljavam za/*na/*∅ provala na sreštata. regret-1s for/ of/ failure-the of meeting-the “I’m sorry about the failure of the meeting.”

• Another property of pu-clauses shared with deto-clauses: they cannot appear in subject position (Bulgarian cited from Rudin 1986: 45): (22) a. *[Pu efije] me stenoxorese that left-3s me upset-3s “That he left, upset me.” b. *[Deto toj e xubav], ne me iznenada that he is handsome not me surprised-3s “That he is handsome, did not surprise me.”

In Greek, (22a) is OK with oti preceded by the definite (neuter) article to: to oti efije me stenoxorese (a nominalized clause; recall that pu-clauses resist nominalization).

How do these properties relate with the structure and islandhood of pu-complements?

• Krapova (2010): deto clauses are embedded under an abstract tova ‘this’ which is a prepositional object, i.e.:

(23) săžaljavam [PP za/∅ [DP tova/∅ [CP deto …. ]]]

With deto, both the preposition and tova remain unrealized.

[6]

• Do we need an abstract complex structure of this sort? If pu-clauses, for example, behave like oblique arguments, does this mean that they are embedded under an abstract P-DP? • If that is the case, why are P and D null and never overtly realized when pu is present? • The structure in (23) may assimilate the pu-complement to a relative modifying ‘this’. But it’s all abstract.

• A different suggestion (a) What we see is what we get, i.e. no abstract unrealized heads. (b) The element pu merges directly with the selecting predicate (see Manzini (2010) on the point of Merge of complementizers, as (nominal) elements outside the embedded IP). (c) Complementizers like oti merge in the ‘canonical’ (accusative) position of internal arguments. (d) Pu merges in the position retained for dative or locative arguments (including indirect objects). In this sense it introduces a clause which behaves like an oblique argument in terms of extraction.

(24) a. *Tinosi stenoxorjese me/jia tis praksis ti? whose be.upset-2s for/with the actions “About/for whose actions are you upset?” b. Me/jia tinos tis praksis stenoxorjese? c. Tinosi akuses tin foni ti?whose heard-2s the voice “Whose voice did you hear?”

• Why ‘obliques’ create islands is a separate issue. Postulating extra functional structure (or sequences as in Caha (2008)) offers a differ description of the problem, without really offering an explanatory account (the idea being that heavier structures in features/functional heads, create more opaque domains/interveners).

3.3 The structure of pu-clauses: Complements or relatives? • An answer to this question partly depends on: a) on how one analyzes relatives (the raising vs. matching/null operator account), b) whether there is a structural distinction between adjuncts and complements.

• A matching analysis may require a null operator, plus adjunction. • A raising analysis (like Kayne’s 1994) requires neither, but may have different shortcomings.

(25) a. Sinandisa [D ton [N fititi [C pu [I pire to vravio ]]]]. “I met the student that got the prize.”

3 Sinandisa(acc) 3 ton 3 fititi 3 pu I(P) 6 pire (pro) to vravio

[7]

b. Stenoxorjeme pu efijes “I’m upset that you left.”

3 stenoxorjeme 3 pu(oblique) IP

In relative clauses there has to be some predication relation between the head of the relative and the clause headed by pu.

4. In short • The same complementizer in relatives, complements, and as an interrogative pronoun. • Pu corresponds to a position identified with oblique arguments. • Whether complement clauses are hidden relatives or vice versa depends on the assumptions we start with regarding these two constructions. Crucially though viewing complementizers as pronominals can provide us with a fresher look of the properties of these constructions.

References Alexopoulou, Th. 2006. Resumption in relative clauses. Natural Language and Linguistic Theory 24: 57-111. Caha, P. 2008. The case hierarchy as functional sequence. In: M. Richards & A. L. Malchukov, Linguistische Berichte 86: 247-276, Universität Leipzig. de Cuba, C. & B. Ürögdi. 2010. Clearing up the ‘facts’ in complementation. UPenn Working Papers in Linguistics 16: 41-50. Available on line: http://repository.upenn.edu/pwpl/vol16/iss1/6 Haegeman, L. & B. Ürögdi. 2010. Referential CPs and DPs: An operator movement account. Theoretical Linguistics 36-2/3: 111-151. Kalluli, D. 2008. Clitic doubling, agreement and information structure. In: Kallulli, D., 2008. Clitic doubling, agreement and information structure, pp. 227-256. In: Kallulli, D., Tasmowski, L. (Eds.), Clitic Doubling in the Balkan Languages. John Benjamins, Amsterdam. Kayne, R. 1994. The antisymmetry of syntax. MIT Press: Cambridge, MA. Kiparsky, P. & C. Kiparsky. 1971. Fact. In: D. D. Steinberg and L. A. Jakobovits (Eds.), Semantics: An Interdisciplinary Reader in Philosophy, Linguistics and Psychology, pp. 345-369. Cambridge University Press: Cambridge. Krapova, I. 2010. Bulgarian relative and factive clauses with an invariant complementizer. Lingua 120: 1240- 1272. Legate, J.A. 2010. On how how is used instead of that. Natural Language and Linguistic Theory 28: 121-134. Manzini, M.R. & L.M. Savoia. 2003. The nature of complementizers. Rivista di Grammatica Generativa 28, 87- 110. Manzini, M.R. 2010. The structure and interpretation of (Romance) complementizers. In: Ph. Panagiotidis (ed.), The complementizer phase, pp. 167-199. Oxford University Press: Oxford. Nicholas, N. 1998. The story of ‘pu’. The grammaticalization in space and time of a Modern Greek complementizer. PhD dissertation, University of Melbourne. Available online at: http://www.tlg.uci.edu/~opoudjis/Work/thesis.html Pesetsky, D. 1995. Zero Syntax. Experiencers and cascades. MIT Press: Cambridge, MA. Roberts, I. & A. Roussou. 2003. Syntactic change. A minimalist approach to grammaticalization. Cambridge University Press: Cambridge. Roussou, A. & I. Roberts. 2001. Pou-complements and Acc-ing constructions: a comparative analysis. Proceedings of the 4th International Conference on Greek Linguistics, Nicosia, September 1999, pp. 201- 208. University Studio Press: Thessaloniki. Roussou, A. 1994. The syntax of complementisers. PhD dissertation, University College London. Rudin, C. 1986. Aspects of Bulgarian syntax: Complementizers and wh-constructions. Slavica Publishers: Columbus, Ohio. Theophanopoulou-Kontou, D. 1985. Οι που-αναφορικές προτάσεις της Νέας Ελληνικής και η παράλειψη της προθέσεως. Μελέτες για την Ελληνική Γλώσσα 2: 51-78. Varlokosta, S. 1994. Issues on Modern Greek sentential complementation. PhD dissertation, University of Maryland, College Park.

[8]